Prediction of protein-protein binding free energies

We present an energy function for predicting binding free energies of protein-protein complexes, using the three-dimensional structures of the complex and unbound proteins as input. Our function is a linear combination of nine terms and achieves a correlation coefficient of 0.63 with experimental measurements when tested on a benchmark of 144 complexes using leave-one-out cross validation. Although we systematically tested both atomic and residue-based scoring functions, the selected function is dominated by residue-based terms. Our function is stable for subsets of the benchmark stratified by experimental pH and extent of conformational change upon complex formation, with correlation coefficients ranging from 0.61 to 0.66.     

A survey of hemoglobin quaternary structures

We perform an analysis of the quaternary structure and dimer/dimer interface in the crystal structures of 165 human hemoglobin tetramers; 112 are in the T, 17 the R, 14 the Y (or R2) state; 11 are high-affinity T state mutants, and 11 may either be intermediates between the states, or off the allosteric transition pathway. The tertiary structure is fixed within each state, in spite of the different ligands, mutations, and chemical modifications present in individual entries. The geometry of the tetramer assembly is essentially the same in all the R or the Y state entries; it is slightly different in high salt and low salt crystals of T state hemoglobins. The dimer/dimer interface differs in terms of size, chemical composition and polar interactions, between the states. It is loosely packed, like crystal packing contacts or the subunit interface of weakly associated homodimers, and unlike most oligomeric proteins, which have close-packed interfaces. The loose packing is most obvious in the liganded forms, where the tetramer is known to dissociate at low concentration. We identify cavities that contribute to the loose packing of the α1β2 and α2β1 contacts. Two pairs of cavities occur recurrently in both the T and the R state tetramers. They may contribute to the allosteric mechanism by facilitating the subunit movements and the tertiary structure changes that accompany the transition from T to R to Y.     

Scoring optimisation of unbound protein-protein docking including protein binding site predictions

The prediction of the structure of the protein-protein complex is of great importance to better understand molecular recognition processes. During systematic protein-protein docking, the surface of a protein molecule is scanned for putative binding sites of a partner protein. The possibility to include external data based on either experiments or bioinformatic predictions on putative binding sites during docking has been systematically explored. The external data were included during docking with a coarse-grained protein model and on the basis of force field weights to bias the docking search towards a predicted or known binding region. The approach was tested on a large set of protein partners in unbound conformations. The significant improvement of the docking performance was found if reliable data on the native binding sites were available. This was possible even if data for single key amino acids at a binding interface are included. In case of binding site predictions with limited accuracy, only modest improvement compared with unbiased docking was found. The optimisation of the protocol to bias the search towards predicted binding sites was found to further improve the docking performance resulting in approximately 40% acceptable solutions within the top 10 docking predictions compared with 22% in case of unbiased docking of unbound protein structures.     

A knowledge-based forcefield for protein-protein interface design

A distance-dependent knowledge-based potential for protein-protein interactions is derived and tested for application in protein design. Information on residue type specific C(alpha) and C(beta) pair distances is extracted from complex crystal structures in the Protein Data Bank and used in the form of radial distribution functions. The use of only backbone and C(beta) position information allows generation of relative protein-protein orientation poses with minimal sidechain information. Further coarse-graining can be done simply in the same theoretical framework to give potentials for residues of known type interacting with unknown type, as in a one-sided interface design problem. Both interface design via pose generation followed by sidechain repacking and localized protein-protein docking tests are performed on 39 nonredundant antibody-antigen complexes for which crystal structures are available. As reference, Lennard-Jones potentials, unspecific for residue type and biasing toward varying degrees of residue pair separation are used as controls. For interface design, the knowledge-based potentials give the best combination of consistently designable poses, low RMSD to the known structure, and more tightly bound interfaces with no added computational cost. 77% of the poses could be designed to give complexes with negative free energies of binding. Generally, larger interface separation promotes designability, but weakens the binding of the resulting designs. A localized docking test shows that the knowledge-based nature of the potentials improves performance and compares respectably with more sophisticated all-atoms potentials.     

A protein-RNA docking benchmark (I): nonredundant cases

We have developed a nonredundant protein-RNA docking benchmark dataset, which is derived from the available bound and unbound structures in the Protein Data Bank involving polypeptide and nucleic acid chains. It consists of nine unbound-unbound cases where both the protein and the RNA are available in the free form. The other 36 cases are of unbound-bound type where only the protein is available in the free form. The conformational change upon complex formation is calculated by a distance matrix alignment method, and based on that, complexes are classified into rigid, semi-flexible, and full flexible. Although in the rigid body category, no significant conformational change accompanies complex formation, the fully flexible test cases show large domain movements, RNA base flips, etc. The benchmark covers four major groups of RNA, namely, t-RNA, ribosomal RNA, duplex RNA, and single-stranded RNA. We find that RNA is generally more flexible than the protein in the complexes, and the interface region is as flexible as the molecule as a whole. The structural diversity of the complexes in the benchmark set should provide a common ground for the development and comparison of the protein-RNA docking methods. The benchmark can be freely downloaded from the internet.     

Apolipoprotein E-low density lipoprotein receptor binding: study of protein-protein interaction in rationally selected docked complexes

Apolipoprotein E (apoE) is an important protein involved in lipid metabolism due to its interaction with members of the low-density lipoprotein receptor (LDLR) family. To further understand the molecular basis for this receptor-binding activity, an apoE fragment containing the receptor binding region (residues 135-151) was docked onto the fifth LDLR ligand binding repeat (LR5) by computational methods. A subset of structures generated by the docking was rationally selected on the grounds of experimental data combined with modeling and was used for further analysis. The application and comparison of two different experimental structures for the apoE fragment underlines the local structural changes occurring in apoE when switching from a receptor-inactive to a receptor-active conformation. The body of interactions occurring at the interface between the two proteins is in very good agreement with the biochemical data available for both apoE and LDLR. Charged residues are involved in numerous ionic interactions and might therefore be important for the specificity of the interaction between apoE and LR5. In addition, the interface also features a tryptophan and a stacking of histidine residues, revealing that the association between the two proteins is not entirely governed by ionic interactions. In particular, the presence of histidine residues in the interface gives a structural basis for the pH-regulated release mechanism of apoE in the endosomes. The proposed molecular basis for apoE binding to LDLR could aid the design of strategies for targeting alterations in lipid transport and metabolism.     

An iterative knowledge-based scoring function for protein-protein recognition

Using an efficient iterative method, we have developed a distance-dependent knowledge-based scoring function to predict protein-protein interactions. The function, referred to as ITScore-PP, was derived using the crystal structures of a training set of 851 protein-protein dimeric complexes containing true biological interfaces. The key idea of the iterative method for deriving ITScore-PP is to improve the interatomic pair potentials by iteration, until the pair potentials can distinguish true binding modes from decoy modes for the protein-protein complexes in the training set. The iterative method circumvents the challenging reference state problem in deriving knowledge-based potentials. The derived scoring function was used to evaluate the ligand orientations generated by ZDOCK 2.1 and the native ligand structures on a diverse set of 91 protein-protein complexes. For the bound test cases, ITScore-PP yielded a success rate of 98.9% if the top 10 ranked orientations were considered. For the more realistic unbound test cases, the corresponding success rate was 40.7%. Furthermore, for faster orientational sampling purpose, several residue-level knowledge-based scoring functions were also derived following the similar iterative procedure. Among them, the scoring function that uses the side-chain center of mass (SCM) to represent a residue, referred to as ITScore-PP(SCM), showed the best performance and yielded success rates of 71.4% and 30.8% for the bound and unbound cases, respectively, when the top 10 orientations were considered. ITScore-PP was further tested using two other published protein-protein docking decoy sets, the ZDOCK decoy set and the RosettaDock decoy set. In addition to binding mode prediction, the binding scores predicted by ITScore-PP also correlated well with the experimentally determined binding affinities, yielding a correlation coefficient of R = 0.71 on a test set of 74 protein-protein complexes with known affinities. ITScore-PP is computationally efficient. The average run time for ITScore-PP was about 0.03 second per orientation (including optimization) on a personal computer with 3.2 GHz Pentium IV CPU and 3.0 GB RAM. The computational speed of ITScore-PP(SCM) is about an order of magnitude faster than that of ITScore-PP. ITScore-PP and/or ITScore-PP(SCM) can be combined with efficient protein docking software to study protein-protein recognition.     

Rotamer libraries and probabilities of transition between rotamers for the side chains in protein-protein binding

Conformational changes in the side chains are essential for protein-protein binding. Rotameric states and unbound- to-bound conformational changes in the surface residues were systematically studied on a representative set of protein complexes. The side-chain conformations were mapped onto dihedral angles space. The variable threshold algorithm was developed to cluster the dihedral angle distributions and to derive rotamers, defined as the most probable conformation in a cluster. Six rotamer libraries were generated: full surface, surface noninterface, and surface interface-each for bound and unbound states. The libraries were used to calculate the probabilities of the rotamer transitions upon binding. The stability of amino acids was quantified based on the transition maps. The noninterface residues' stability was higher than that of the interface. Long side chains with three or four dihedral angles were less stable than the shorter ones. The transitions between the rotamers at the interface occurred more frequently than on the noninterface surface. Most side chains changed conformation within the same rotamer or moved to an adjacent rotamer. The highest percentage of the transitions was observed primarily between the two most occupied rotamers. The probability of the transition between rotamers increased with the decrease of the rotamer stability. The analysis revealed characteristics of the surface side-chain conformational transitions that can be utilized in flexible docking protocols.     

Dissecting protein-protein recognition sites

The recognition sites in 70 pairwise protein-protein complexes of known three-dimensional structure are dissected in a set of surface patches by clustering atoms at the interface. When the interface buries <2000 A2 of protein surface, the recognition sites usually form a single patch on the surface of each component protein. In contrast, larger interfaces are generally multipatch, with at least one pair of patches that are equivalent in size to a single-patch interface. Each recognition site, or patch within a site, contains a core made of buried interface atoms, surrounded by a rim of atoms that remain accessible to solvent in the complex. A simple geometric model reproduces the number and distribution of atoms within a patch. The rim is similar in composition to the rest of the protein surface, but the core has a distinctive amino acid composition, which may help in identifying potential protein recognition sites on single proteins of known structures.     

Interaction geometry involving planar groups in protein-protein interfaces

The geometry of interactions of planar residues is nonrandom in protein tertiary structures and gives rise to conventional, as well as nonconventional (X--H...pi, X--H...O, where X = C, N, or O) hydrogen bonds. Whether a similar geometry is maintained when the interaction is across the protein-protein interface is addressed here. The relative geometries of interactions involving planar residues, and the percentage of contacts giving rise to different types of hydrogen bonds are quite similar in protein structures and the biological interfaces formed by protein chains in homodimers and protein-protein heterocomplexes--thus pointing to the similarity of chemical interactions that occurs during protein folding and binding. However, the percentage is considerably smaller in the nonspecific and nonphysiological interfaces that are formed in crystal lattices of monomeric proteins. The C--H...O interaction linking the aromatic and the peptide groups is quite common in protein structures as well as the three types of interfaces. However, as the interfaces formed by crystal contacts are depleted in aromatic residues, the weaker hydrogen bond interactions would contribute less toward their stability.     

Potential of mean force for protein-protein interaction studies.

Calculating protein-protein interaction energies is crucial for understanding protein-protein associations. On the basis of the methodology of mean-field potential, we have developed an empirical approach to estimate binding free energy for protein-protein interactions. This knowledge-based approach has been used to derive distance-dependent free energies of protein complexes from a nonredundant training set in the Protein Data Bank (PDB), with a careful treatment of homology. We calculate atom pair potentials for 16 pair interactions, which can reflect the importance of hydrophobic interactions and specific hydrogen-bonding interactions. The derived potentials for hydrogen-bonding interactions show a valley of favorable interactions at a distance of approximately 3 A, corresponding to that of an established hydrogen bond. For the test set of 28 protein complexes, the calculated energies have a correlation coefficient of 0.75 compared with experimental binding free energies. The performance of the method in ranking the binding energies of different protein-protein complexes shows that the energy estimation can be applied to value binding free energies for protein-protein associations.     

Reassessing buried surface areas in protein–protein complexes

The buried surface area (BSA), which measures the size of the interface in a protein–protein complex may differ from the accessible surface area (ASA) lost upon association (which we call DSA), if conformation changes take place. To evaluate the DSA, we measure the ASA of the interface atoms in the bound and unbound states of the components of 144 protein–protein complexes taken from the Protein–Protein Interaction Affinity Database of Kastritis et al. (2011). We observe differences exceeding 20%, and a systematic bias in the distribution. On average, the ASA calculated in the bound state of the components is 3.3% greater than in their unbound state, and the BSA, 7% greater than the DSA. The bias is observed even in complexes where the conformation changes are small. An examination of the bound and unbound structures points to a possible origin: local movements optimize contacts with the other component at the cost of internal contacts, and presumably also the binding free energy.     

Combining different design strategies for rational affinity maturation of the MICA-NKG2D interface

We redesigned residues on the surface of MICA, a protein that binds the homodimeric immunoreceptor NKG2D, to increase binding affinity with a series of rational, incremental changes. A fixed-backbone RosettaDesign protocol scored a set of initial mutations, which we tested by surface plasmon resonance for thermodynamics and kinetics of NKG2D binding, both singly and in combination. We combined the best four mutations at the surface with three affinity-enhancing mutations below the binding interface found with a previous design strategy. After curating design scores with three cross-validated tests, we found a linear relationship between free energy of binding and design score, and to a lesser extent, enthalpy and design score. Multiple mutants bound with substantial subadditivity, but in at least one case full additivity was observed when combining distant mutations. Altogether, combining the best mutations from the two strategies into a septuple mutant enhanced affinity by 50-fold, to 50 nM, demonstrating a simple, effective protocol for affinity enhancement.     

A minimal model of protein–protein binding affinities

A minimal model of protein–protein binding affinity that takes into account only two structural features of the complex, the size of its interface, and the amplitude of the conformation change between the free and bound subunits, is tested on the 144 complexes of a structure‐affinity benchmark. It yields K_d values that are within two orders of magnitude of the experiment for 67% of the complexes, within three orders for 88%, and fails on 12%, which display either large conformation changes, or a very high or a low affinity. The minimal model lacks the specificity and accuracy needed to make useful affinity predictions, but it should help in assessing the added value of parameters used by more elaborate models, and set a baseline for evaluating their performances.     

Effects of point mutations in pVHL on the binding of HIF-1α

The von Hippel-Lindau tumor suppressor protein (pVHL) has an essential role in the regulation of the hypoxia response pathway in animal cells. Under normoxic conditions, the hypoxia-inducible factor (HIF) undergoes trans-4-prolyl hydroxylation and is subsequently recognised by the β-domain of pVHL, leading to the ubiquitination and degradation of HIF. Mutations of pVHL alter the binding of HIF. A subset of relevant clinically observed mutations to pVHL are thought to cause weaker binding of HIF-1α and are associated with cancer and cardiovascular diseases. Here, we present computational studies analyzing the interaction of HIF with mutant forms of pVHL, describing at atomic detail the local structural reorganization caused by substitution of certain residues of pVHL. The results reveal that the canonical configuration in the wild-type system is vital for the efficient functioning of the complex and that mutation of any of the residues implicated in the h-bond network in the binding site disrupts HIF binding. Although the experimentally observed ordering of binding energies for mutants of Tyr98 is reproduced, our examination of a broader range of mutations does not support the hypothesis of a correlation between the degree of disruption of the pVHL/HIF-1α interaction caused by a mutation and the phenotype with which the mutation is associated. We suggest that disruption of the binding interaction is one of many factors behind the manifestation of VHL disease.     

Assessing predictions of protein-protein interaction: the CAPRI experiment

The Critical Assessment of PRedicted Interactions (CAPRI) experiment was designed in 2000 to test protein docking algorithms in blind predictions of the structure of protein-protein complexes. In four years, 17 complexes offered by crystallographers as targets prior to publication, have been subjected to structure prediction by docking their two components. Models of these complexes were submitted by predictor groups and assessed by comparing their geometry to the X-ray structure and by evaluating the quality of the prediction of the regions of interaction and of the pair wise residue contacts. Prediction was successful on 12 of the 17 targets, most of the failures being due to large conformation changes that the algorithms could not cope with. Progress in the prediction quality observed in four years indicates that the experiment is a powerful incentive to develop new procedures that allow for flexibility during docking and incorporate nonstructural information. We therefore call upon structural biologists who study protein-protein complexes to provide targets for further rounds of CAPRI predictions.     

The dataset for protein–RNA binding affinity

We have developed a non‐redundant protein–RNA binding benchmark dataset derived from the available protein–RNA structures in the Protein Database Bank. It consists of 73 complexes with measured binding affinity. The experimental conditions (pH and temperature) for binding affinity measurements are also listed in our dataset. This binding affinity dataset can be used to compare and develop protein–RNA scoring functions. The predicted binding free energy of the 73 complexes from three available scoring functions for protein–RNA docking has a low correlation with the binding Gibbs free energy calculated from K_d. © 2013 The Protein Society     

Calculating proton uptake/release and binding free energy taking into account ionization and conformation changes induced by protein-inhibitor association: application to plasmepsin, cathepsin D and endothiapepsin-pepstatin complexes

The protein-inhibitor binding energies of enzymes are often pH dependent, and binding induces either proton uptake or proton release. The proton uptake/release and the binding energy for three complexes with available experimental data were numerically studied: pepstatin-cathepsin D, pepstatin-plasmepsin II and pepstatin-endothiapepsin. Very good agreement with the experimental data was achieved when conformational changes were taken into account. The role of the desolvation energy and the conformational changes was revealed by modeling the complex, the separated molecules in the complex conformation and the free molecules. It was shown that the conformational changes induced by the complex formation are as important for the proton transfer as the loss of solvation energy caused by the burial of interface residues. The residues responsible for the proton transfer were identified and their contribution to the proton uptake/release calculated. These residues were found to be scattered along the whole protein rather than being localized only at the active site. In the case of cathepsin D, these residues were found to be highly conserved among the cathepsin D sequences of other species. It was shown that conformation and ionization changes induced by the complex formation are critical for the correct calculation of the binding energy. Taking into account the electrostatics and the van der Waals (vdW) energies within the Boltzmann distribution of energies and allowing ionization and conformation changes to occur makes the calculated binding energy more realistic and closer to the experimental value. The interplay between electrostatic and vdW forces makes the pH dependence of the binding energy smoother, because the vdW force acts in reaction to the changes of the electrostatic energy. It was found that a small fraction of the ionizable groups remain uncharged in both the free and complexed molecules. The sequence and structural position of these groups aligns well within the three proteases, suggesting that these may have specific role.     

Properties of polyproline II, a secondary structure element implicated in protein-protein interactions

The polyproline II (PPII) conformation of protein backbone is an important secondary structure type. It is unusual in that, due to steric constraints, its main-chain hydrogen-bond donors and acceptors cannot easily be satisfied. It is unable to make local hydrogen bonds, in a manner similar to that of alpha-helices, and it cannot easily satisfy the hydrogen-bonding potential of neighboring residues in polyproline conformation in a manner analogous to beta-strands. Here we describe an analysis of polyproline conformations using the HOMSTRAD database of structurally aligned proteins. This allows us not only to determine amino acid propensities from a much larger database than previously but also to investigate conservation of amino acids in polyproline conformations, and the conservation of the conformation itself. Although proline is common in polyproline helices, helices without proline represent 46% of the total. No other amino acid appears to be greatly preferred; glycine and aromatic amino acids have low propensities for PPII. Accordingly, the hydrogen-bonding potential of PPII main-chain is mainly satisfied by water molecules and by other parts of the main-chain. Side-chain to main-chain interactions are mostly nonlocal. Interestingly, the increased number of nonsatisfied H-bond donors and acceptors (as compared with alpha-helices and beta-strands) makes PPII conformers well suited to take part in protein-protein interactions.